RFID pose estimation non-line-of-sight shielding influence suppression method based on antenna weight
By constructing a tag position and non-line-of-sight occlusion error model and antenna weights, and combining augmented Lagrange convex optimization and multi-signal classification algorithms, the problem of non-line-of-sight occlusion influence in RFID pose estimation is solved, improving estimation accuracy and stability, and making it suitable for complex industrial scenarios.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HEBEI UNIV OF TECH
- Filing Date
- 2026-01-23
- Publication Date
- 2026-05-05
AI Technical Summary
In complex industrial scenarios, RFID pose estimation is affected by non-line-of-sight occlusion, which leads to signal parameter distortion and introduces systematic bias, making it difficult to achieve accurate and stable pose estimation. In particular, the accumulation of errors is severe in dynamic positioning scenarios, affecting estimation accuracy and real-time performance.
A model relating tag position to non-line-of-sight occlusion error is constructed, antenna weights are assigned, and the effects of non-line-of-sight occlusion are suppressed by augmented Lagrange convexity optimization algorithm and multi-signal classification algorithm, so as to achieve synchronous estimation of tag position and attitude information.
It improves the pose estimation accuracy of RFID in non-line-of-sight environments, suppresses error accumulation, achieves continuous and stable pose tracking, and simultaneously estimates position and attitude information. It does not require additional hardware and is suitable for indoor industrial scenarios with limited space.
Smart Images
Figure CN121980313A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of wireless sensing technology, specifically to a method for suppressing non-line-of-sight occlusion effects in RFID pose estimation based on antenna weights. Background Technology
[0002] Accurate acquisition of object position information in industrial settings is crucial for the informatization and intelligentization of industrial processes. Industrial Internet of Things (IIoT) technology can integrate information such as the object's environment and identity, using object position information as a carrier. Furthermore, IIoT technology can correlate the location information of objects and production units, expanding the sensing range of production modules. This integration of information and expansion of the sensing range effectively improves production line efficiency. Therefore, indoor positioning technology based on industrial scenarios has become a highly anticipated new technology in the fields of IIoT and intelligent manufacturing.
[0003] Some traditional indoor positioning technologies are limited by various factors, making it difficult to accurately estimate the pose of objects in complex industrial scenarios. For example, visual positioning technology requires high computational power and sophisticated equipment, and is heavily dependent on lighting conditions, making it difficult to achieve accurate object tracking and positioning in industrial scenarios with significant non-line-of-sight obstruction and complex lighting distribution. Inertial navigation positioning technology generates unavoidable offsets and cumulative errors during the positioning process, making it difficult to meet the long-term positioning requirements in industrial scenarios. Compared to the above two methods, radio wave positioning technology, due to its low computational complexity, strong environmental adaptability, and non-contact multi-target positioning characteristics, is more suitable for indoor positioning in industrial scenarios. Although RFID-based indoor positioning technology has made significant progress driven by emerging technologies such as smart manufacturing and the Internet of Things;
[0004] Indoor industrial settings are typically space-constrained, compact, and involve frequent personnel movement and complex machine operations, making them prone to non-line-of-sight (NFS) occlusion. Under such occlusion, radio frequency (RF) signals experience severe attenuation, leading to distortion of signal parameters acquired by the receiver. Pose calculations based on this distorted signal introduce significant systematic bias, drastically reducing pose estimation accuracy. In dynamic positioning scenarios, NFS errors accumulate and amplify during continuous pose estimation of moving objects, resulting in a continuous decline in positioning accuracy and failing to meet the requirements for continuous and stable pose estimation in industrial settings. Furthermore, due to the inherent characteristics of RFID technology, RFID signals contain parameters such as target attitude, position, and environmental parameters, which are strongly coupled within the signal model. Traditional methods struggle to achieve joint decoupling of multiple parameters within a single observation framework, often requiring independent estimation. This leads to asynchronous and mutually restrictive acquisition of position and attitude information, impacting the overall performance and real-time nature of pose estimation. Therefore, we propose an antenna weight-based method for suppressing the impact of NFS occlusion in RFID pose estimation. Summary of the Invention
[0005] The purpose of this invention is to provide a method for suppressing non-line-of-sight occlusion effects in RFID pose estimation based on antenna weights, so as to solve the problems mentioned in the background art.
[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for suppressing non-line-of-sight occlusion effects in RFID pose estimation based on antenna weights, the method comprising the following steps:
[0007] S1. Construct a model relating label position to non-line-of-sight occlusion error;
[0008] S2. Assign antenna weights based on the theoretical and actual phase differences between each antenna and the reference antenna;
[0009] S3. Based on the relationship model of S1 and the antenna weights of S2, an augmented Lagrange convex optimization algorithm is constructed. By continuously iterating to suppress the influence of non-line-of-sight occlusion, the tag location information is finally obtained.
[0010] S4. Combining antenna weights and multi-signal classification algorithms, a weighted covariance matrix is constructed to suppress the influence of non-line-of-sight occlusion, and finally, the tag attitude information is obtained through peak detection.
[0011] Optionally, step S1 includes the following steps:
[0012] The phase information of the tag is obtained through the reader, and the phase value φ can be expressed as:
[0013]
[0014] In the formula, d represents the distance between the antenna and the tag, and φ h φ p φ t These represent the phase shifts caused by the reader circuit, tag orientation, and tag hardware, respectively; f is the antenna's transmission frequency; and c is the speed of light.
[0015] The first antenna is placed in a line-of-sight environment to serve as the reference antenna, while the remaining antennas are within a preset recognition range. The distance φ is expanded to obtain the measured distance D from the i-th antenna to the tag. i Then the difference between the squared distances of the i-th antenna and the reference antenna can be expressed as:
[0016]
[0017] In the formula, (x,y) represents the label coordinates, (x...y ... i ,y i ) represents the coordinates of the i-th antenna, φ i,nlos d represents the phase error of the i-th antenna caused by non-line-of-sight obstruction. i,nlos d represents the distance error caused by non-line-of-sight obstruction of the i-th antenna.i L represents the actual distance from the i-th antenna to the tag. i =x2 i + y2 i;
[0018] Define D i,1 =D i -D1, the relationship between label position and non-line-of-sight occlusion error is:
[0019]
[0020] The system has N antennas. The relationship between the tag position τ and the non-line-of-sight occlusion error is modeled as follows:
[0021]
[0022] In the formula, .
[0023] Optionally, step S2 includes the following steps:
[0024] Based on the received phase values of each antenna, the initial coarse estimated coordinates of the tag are obtained using the least squares method. Since the antenna positions are fixed and known, the theoretical phase difference Δθ between the i-th antenna and the reference antenna can be calculated using the initial coarse estimated coordinates of the tag. i,1 The difference between the theoretical phase difference and the actual phase difference is calculated to obtain the phase difference offset value ΔΦ. i,1 Based on the distribution of phase difference offset values P(ΔΦ) i,1 |τ) Assign antenna weights ω:
[0025]
[0026] In the formula, P(ΔΦ) i,1 The numerator of |τ) is the Gaussian probability formula, and the denominator is the normalization coefficient, μ. i,1 and σ i,1 These are the mean and standard values of the phase measurement residuals for the i-th antenna.
[0027] Optionally, step S3 includes the following steps:
[0028] Combining the relational model and antenna weights, we can obtain the weighted relational model α:
[0029]
[0030] Taking the logarithm of the Gaussian probability density function P(α|τ) for α yields the objective function F. τ :
[0031]
[0032] In the formula, u iLet τ be the mean value of the phase measured by the i-th antenna. This function is convex, and the probability of the actual error distribution matching the theoretical error distribution is maximized when it reaches its minimum value. At this point, τ represents the ideal coordinates of the tag.
[0033] Based on the convex function in the above equation, an augmented Lagrange convex optimization algorithm is constructed:
[0034]
[0035] In the formula, λ is the Lagrange multiplier, s is the slack variable, μ is the penalty parameter, and h τ Let X be the constraint vector. max Y max These represent the maximum x-coordinate and y-coordinate of the measurement scene, respectively. The algorithm is continuously updated and iterated to finally obtain the ideal coordinate value τ of the label.
[0036] Optionally, the S4 process is as follows:
[0037] For the i-th antenna, perform differential operations on the phase signals obtained from adjacent tags in the same acquisition to obtain a phase signal matrix x that does not contain position parameters. i (k):
[0038]
[0039] The signal covariance matrix R is obtained from the signal matrix. i :
[0040] (10)
[0041] Construct a weighted covariance matrix R by combining antenna weights ω And obtain its eigenvector R;
[0042]
[0043] In the formula, U N It is R ω The noise subspace composed of smaller eigenvalues, U s It is R ω The signal subspace composed of larger eigenvalues;
[0044] Based on the weighted covariance matrix, the multi-signal classification spectrum P is obtained. MUSIC ;
[0045] ;
[0046] a ω (θ) is related to U N Orthogonal direction vectors, P MUSIC The peak point is the tag attitude angle.
[0047] Compared with existing technologies, this invention provides a method for suppressing non-line-of-sight occlusion effects in RFID pose estimation based on antenna weights, which has the following beneficial effects:
[0048] 1. This antenna weight-based RFID pose estimation non-line-of-sight occlusion suppression method effectively improves the pose estimation accuracy of RFID in non-line-of-sight environments. By constructing a relationship model between tag position and non-line-of-sight occlusion error, the error magnitude is quantified, rather than simply treating it as environmental noise. In addition, through an adaptive antenna weight allocation mechanism based on phase residual statistical analysis, antenna occlusion is automatically identified and the corresponding weight is reduced. The model and weights are integrated into an augmented Lagrange convex optimization function with physical space constraints to achieve target position estimation. The weights are integrated into a multi-signal classification algorithm to achieve target pose estimation. This method can maintain high reliability and robustness in complex industrial scenarios, thus improving the pose estimation accuracy of RFID in non-line-of-sight environments.
[0049] 2. This antenna weight-based RFID pose estimation non-line-of-sight occlusion suppression method effectively suppresses error accumulation in dynamic scenes and achieves continuous and stable pose tracking. For moving targets, the non-line-of-sight error in traditional methods accumulates over time, leading to tracking failure. In the estimation at each moment, the non-line-of-sight error is actively estimated and compensated through function optimization, rather than being passively ignored or smoothed. This effectively suppresses the propagation and amplification of errors, ensuring the reliability of continuous and stable pose tracking of moving objects.
[0050] 3. This antenna weight-based RFID pose estimation non-line-of-sight occlusion suppression method achieves synchronous estimation of position and attitude information, improving pose estimation efficiency. Adaptive antenna weights are applied simultaneously to both position and attitude estimation. In the position estimation stage, the weights are used to construct a weighted objective function, eliminating the influence of attitude parameters through tag coordinates. In the attitude estimation stage, the weights are used to construct a weighted covariance matrix, eliminating the influence of position parameters through phase difference. This organically combines the position and attitude estimation processes without interference, achieving synchronous and independent estimation of position and attitude.
[0051] 4. This antenna weight-based RFID pose estimation non-line-of-sight occlusion suppression method uses RFID radio for non-contact pose estimation, which does not damage the measured surface. The hardware system is a basic RFID positioning system, requiring no additional hardware. The system structure is simple and easy to deploy in space-constrained indoor industrial settings. Attached Figure Description
[0052] Figure 1 This is a schematic diagram of the pose estimation algorithm of the present invention.
[0053] Figure 2 This is a location diagram of the RFID system of the present invention;
[0054] Figure 3 This is a schematic diagram illustrating the principle of the weighted multi-signal classification algorithm of the present invention;
[0055] Figure 4 This is a diagram showing the location estimation results of the present invention;
[0056] Figure 5 This is a diagram showing the attitude angle estimation results of the present invention. Detailed Implementation
[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0058] In the description of this invention, it should be understood that the orientation descriptions refer only to a two-dimensional plane. Specifically, the x-axis of the two-dimensional coordinate system is defined as the horizontal direction, and the y-axis is defined as the vertical direction.
[0059] This description is provided merely for ease of description and simplification, and is not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it does not limit the scope of protection of this invention. In the description of this invention, unless otherwise expressly limited, technical terms such as suppression, frequency, antenna, and tag should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this invention in conjunction with the specific content of the technical solution.
[0060] like Figures 1-5 As shown, this invention provides a technical solution: a non-line-of-sight occlusion suppression method for RFID pose estimation based on antenna weights. This method utilizes a passive RFID system including a reader, antenna, and RFID tag. The RFID system employing this method can be applied in complex industrial scenarios with non-line-of-sight occlusion, requiring only a basic RFID system without the need for additional peripherals. To better understand this invention, the following description, in conjunction with specific embodiments and accompanying drawings, further illustrates the invention.
[0061] like Figures 1-2 As shown, the component selection and system structure of this invention embodiment are as follows:
[0062] Regarding component selection, the RFID reader is model Impinj R420, configured for dual-frequency transmission with frequencies of 920.625MHz and 924.375MHz; the antenna is model Laird S9028PCL, and the antenna array consists of three antennas of this model, with antenna 1 serving as the reference antenna in a line-of-sight environment, and antennas 2 and 3 in any environment; the tag is model Alien 9654, conforming to the EPC-GEN communication protocol, and a tag array consisting of three tags of the same model is attached to the same plane on the target object; all of the above equipment are conventional commercial equipment and can be obtained commercially.
[0063] In terms of system structure, the target object is placed on the slider of the lead screw slide rail and moves with the slider at a speed of 2cm / s, simulating the moving object on the conveyor belt in an industrial scene; a plane covered with 10-micrometer-thick tin foil is placed between tag 2 and tag 3 and the target object to simulate non-line-of-sight occlusion; in the initial state of the system, antenna 2 and antenna 3 are on the same horizontal plane and 80cm apart; the horizontal distance between antenna 2 and antenna 1 is 80cm and the vertical distance is 230cm; the target object is on the same vertical line as antenna 2, 140cm apart, and the vertical distance between the target object and antenna 1 is 90cm; the effective stroke of the lead screw slide rail is 80cm, and it moves from left to right along the horizontal plane.
[0064] like Figure 1 As shown, the process of this embodiment of the invention includes the following steps:
[0065] S1. Construct a model relating label position to non-line-of-sight occlusion error;
[0066] S2. Assign antenna weights based on the theoretical and actual phase differences between each antenna and the reference antenna;
[0067] S3. Based on the relationship model of S1 and the antenna weights of S2, an augmented Lagrange convex optimization algorithm is constructed. By continuously iterating, the influence of non-line-of-sight occlusion error is suppressed, and finally the tag position information is obtained.
[0068] S4. Combining antenna weights and multi-signal classification algorithms, a weighted covariance matrix is constructed to suppress the influence of non-line-of-sight occlusion, and finally, the tag attitude information is obtained through peak detection.
[0069] Among them, S1 corresponds to the phase deambiguation part in the basic RFID system and data processing algorithm, S2 and S3 correspond to the target position information acquisition part, and S2 and S4 correspond to the target attitude information acquisition part.
[0070] The reader obtains the tag's phase information, and the phase value φ can be represented as:
[0071] In the formula, d represents the distance between the antenna and the tag, and φ h φ p φ t These represent the phase shifts caused by the reader circuit, tag orientation, and tag hardware, respectively, where f is the antenna's transmission frequency and c is the speed of light.
[0072] The first antenna is placed in a line-of-sight environment to serve as the reference antenna, while the remaining antennas are within a preset recognition range. Expanding φ, we obtain the measured distance D from the i-th antenna to the tag. i Then the difference between the squared distances of the i-th antenna and the reference antenna can be expressed as:
[0073]
[0074] In the formula, (x,y) represents the label coordinates, (x...y ... i ,y i ) represents the coordinates of the i-th antenna, φ i,nlos d represents the phase error of the i-th antenna caused by non-line-of-sight obstruction. i,nlos d represents the distance error caused by non-line-of-sight obstruction of the i-th antenna. i L represents the actual distance from the i-th antenna to the tag. i =x²i+y²i.
[0075] Define D i,1 =D i -D1, establishes the relationship between label position and non-line-of-sight occlusion errors:
[0076]
[0077] The system has N antennas. The relationship between the tag position τ and the non-line-of-sight occlusion error is modeled as follows:
[0078]
[0079] In the formula,
[0080] The S2 process is as follows:
[0081] Based on the received phase values of each antenna, the initial coarsely estimated coordinates of the tag are obtained using the least squares method. Since the antenna positions are fixed and known, the theoretical phase difference Δθ between the i-th antenna and the reference antenna can be calculated using the initial coarsely estimated coordinates of the tag. i,1 The phase difference offset value ΔΦ is obtained by calculating the difference between the theoretical phase difference and the actual phase difference. i,1 Based on the distribution of phase difference offset values P(ΔΦ) i,1 |τ) Assign antenna weights ω:
[0082]
[0083] In the formula, P(ΔΦ) i,1 The numerator of |τ) is the Gaussian probability formula, and the denominator is the normalization coefficient, μ. i,1 and σ i,1 These are the mean and standard values of the phase measurement residuals for the i-th antenna.
[0084] The S3 process is as follows:
[0085] Combining the relational model and antenna weights, we can obtain the weighted relational model α:
[0086]
[0087] Taking the logarithm of the Gaussian probability density function P(α|τ) for α yields the objective function F. τ :
[0088]
[0089] In the formula, u i Let τ be the mean value of the phase measured by the i-th antenna. This function is convex, and the probability of the actual error distribution matching the theoretical error distribution is maximized when it reaches its minimum value. At this point, τ is the ideal coordinate value of the tag.
[0090] Based on the convex function in the above equation, an augmented Lagrange convex optimization algorithm is constructed:
[0091]
[0092] In the formula, λ is the Lagrange multiplier, s is the slack variable, μ is the penalty parameter, and h τ X is the constraint vector. max Y max These represent the maximum x-coordinate and y-coordinate of the measurement scene, respectively.
[0093] The algorithm was continuously updated and iterated until the ideal coordinate value τ of the label was finally obtained.
[0094] Furthermore, Figure 3 A schematic diagram illustrating the principle of a multi-signal classification algorithm is shown, such as... Figure 3 As shown, the azimuth angle of the tag array relative to the antenna is θ, the horizontal spacing between the tags is d, and the straight-line distance between the leftmost tag and the antenna is L. Since L≫d, the distances between adjacent tags and the antenna differ by dcosθ. This distance difference causes a change in phase information; therefore, the azimuth angle θ can be estimated by calculating the change in phase information in the RFID signal, i.e., S4.
[0095] The S4 process is as follows:
[0096] Taking the first tag in the tag array as a reference, the phase value of the m-th tag received by the antenna during the k-th measurement is:
[0097]
[0098] Where, φ nlos,m The phase error is caused by non-line-of-sight occlusion.
[0099] Furthermore, for the i-th antenna, the phase signals obtained from adjacent tags in the same acquisition are differentially processed to obtain a phase signal matrix x that does not contain position parameters. i (k):
[0100]
[0101] The signal covariance matrix R is obtained from the signal matrix. i :
[0102]
[0103] Construct a weighted covariance matrix R by combining antenna weights ω And obtain its eigenvector R.
[0104]
[0105] In the formula, U N It is R ω The noise subspace composed of smaller eigenvalues, U s It is R ω The signal subspace composed of larger eigenvalues.
[0106] Based on the weighted covariance matrix, the multi-signal classification spectrum P is obtained. MUSIC :
[0107]
[0108] a ω (θ) is related to U N Orthogonal direction vectors, P MUSIC The peak point is the tag attitude angle.
[0109] like Figure 4 As shown, the positioning results exhibit different errors at different distances. When the vertical distance between the tag and the reference antenna 1 is 120cm, the average total positioning error is 9.84cm. Figure 5 As shown, the pose estimation results exhibit different errors under different tag array attitude angles. When the attitude angle of the tag array relative to the antenna is 30°, the average total attitude estimation error is 2.5°. After the attitude angle is 45°, due to the mutual coupling between tags, the attitude angle estimation error increases significantly, and the estimated value becomes invalid.
[0110] The present invention has been described in detail above. However, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, any modifications or improvements that do not depart from the spirit of the present invention are within the scope of protection of the present invention.
Claims
1. A method for suppressing non-line-of-sight occlusion effects in RFID pose estimation based on antenna weights, characterized in that: The method includes the following steps: S1. Construct a model relating label position to non-line-of-sight occlusion error; S2. Assign antenna weights based on the theoretical and actual phase differences between each antenna and the reference antenna; S3. Based on the relationship model of S1 and the antenna weights of S2, an augmented Lagrange convex optimization algorithm is constructed. By continuously iterating to suppress the influence of non-line-of-sight occlusion, the tag location information is finally obtained. S4. Combining antenna weights and multi-signal classification algorithms, a weighted covariance matrix is constructed to suppress the influence of non-line-of-sight occlusion, and finally, the tag attitude information is obtained through peak detection.
2. The method for suppressing non-line-of-sight occlusion effects in RFID pose estimation based on antenna weights according to claim 1, characterized in that: S1 includes the following steps: The phase information of the tag is obtained through the reader, and the phase value φ can be expressed as: In the formula, d represents the distance between the antenna and the tag, and φ h φ p φ t These represent the phase shifts caused by the reader circuit, tag orientation, and tag hardware, respectively; f is the antenna's transmission frequency; and c is the speed of light. The first antenna is placed in a line-of-sight environment to serve as the reference antenna, while the other antennas are within a preset recognition range; Expanding φ, we obtain the measured distance D from the i-th antenna to the tag. i Then the difference between the squared distances of the i-th antenna and the reference antenna can be expressed as: In the formula, (x,y) represents the label coordinates, (x...y ... i ,y i ) represents the coordinates of the i-th antenna, φ i,nlos d represents the phase error of the i-th antenna caused by non-line-of-sight obstruction. i,nlos d represents the distance error caused by non-line-of-sight obstruction of the i-th antenna. i L represents the actual distance from the i-th antenna to the tag. i =x2 i + y2 i; Define D i,1 =D i -D1, the relationship between label position and non-line-of-sight occlusion error is: The system has N antennas. The relationship between the tag position τ and the non-line-of-sight occlusion error is modeled as follows: In the formula, .
3. The method for suppressing non-line-of-sight occlusion effects in RFID pose estimation based on antenna weights according to claim 1, characterized in that: S2 includes the following steps: Based on the received phase values of each antenna, the initial coarse estimated coordinates of the tag are obtained using the least squares method. Since the antenna positions are fixed and known, the theoretical phase difference Δθ between the i-th antenna and the reference antenna is calculated using the initial coarse estimated coordinates of the tag. i,1 The difference between the theoretical phase difference and the actual phase difference is calculated to obtain the phase difference offset value ΔΦ. i,1 Based on the distribution of phase difference offset values P(ΔΦ) i,1 |τ) Assign antenna weights ω: In the formula, P(ΔΦ) i,1 The numerator of |τ) is the Gaussian probability formula, and the denominator is the normalization coefficient, μ i,1 and σ i,1 These are the mean and standard values of the phase measurement residuals for the i-th antenna.
4. The method for suppressing non-line-of-sight occlusion effects in RFID pose estimation based on antenna weights according to claim 1, characterized in that: S3 includes the following steps: Combining the relational model and antenna weights, we can obtain the weighted relational model α: Taking the logarithm of the Gaussian probability density function P(α|τ) for α yields the objective function F. τ : In the formula, u i Let τ be the mean value of the phase measured by the i-th antenna. This function is convex, and the probability of the actual error distribution matching the theoretical error distribution is maximized when it reaches its minimum value. At this point, τ represents the ideal coordinates of the tag. Based on the convex function in the above equation, an augmented Lagrange convex optimization algorithm is constructed: In the formula, λ is the Lagrange multiplier, s is the slack variable, μ is the penalty parameter, and h τ Let X be the constraint vector. max Y max These represent the maximum x-coordinate and y-coordinate of the measurement scene, respectively.
5. The method for suppressing non-line-of-sight occlusion effects in RFID pose estimation based on antenna weights according to claim 1, characterized in that: The S4 process is as follows: For the i-th antenna, perform differential operations on the phase signals obtained from adjacent tags in the same acquisition to obtain a phase signal matrix x that does not contain position parameters. i (k): The signal covariance matrix R is obtained from the signal matrix. i : (10) Construct a weighted covariance matrix R by combining antenna weights ω And obtain its eigenvector R; In the formula, U N It is R ω The noise subspace composed of smaller eigenvalues, U s It is R ω The signal subspace composed of larger eigenvalues; Based on the weighted covariance matrix, the multi-signal classification spectrum P is obtained. MUSIC ; ； a ω (θ) is related to U N Orthogonal direction vectors, P MUSIC The peak point is the tag attitude angle.